Hybrid Multivalent Jack Bean α-Mannosidase Inhibitors: The First Example of Gold Nanoparticles Decorated with Deoxynojirimycin Inhitopes

Among carbohydrate-processing enzymes, Jack bean α-mannosidase (JBα-man) is the glycosidase with the best responsiveness to the multivalent presentation of iminosugar inhitopes. We report, in this work, the preparation of water dispersible gold nanoparticles simultaneously coated with the iminosugar deoxynojirimycin (DNJ) inhitope and simple monosaccharides (β-d-gluco- or α-d-mannosides). The display of DNJ at the gold surface has been modulated (i) by using an amphiphilic linker longer than the aliphatic chain used for the monosaccharides and (ii) by presenting the inhitope, not only in monomeric form, but also in a trimeric fashion through combination of a dendron approach with glyconanotechnology. The latter strategy resulted in a strong enhancement of the inhibitory activity towards JBα-man, with a Ki in the nanomolar range (Ki = 84 nM), i.e., more than three orders of magnitude higher than the monovalent reference compound.


Introduction
More than ten years have passed since the first example of a trivalent deoxynojirimycin (DNJ) derivative, that displayed a small, but quantifiable, inhibitory multivalent effect for Jack bean α-mannosidase (JBα-man), was reported [1]. Until then, the multivalency concept was considered an exclusive prerogative of lectin-carbohydrate interactions, and its application to carbohydrate-processing enzymes was almost completely unexplored, being considered extremely challenging, from a practical standpoint, and theoretically arguable. The refutation of such speculations was apparent in 2010, when a fullerene-based 12-valent DNJ compound showed binding enhancements towards JBα-man up to three orders of magnitude over the monovalent counterpart (inhibition constant K i = 0.15 µM vs. K i = 188 µM) [2].

Preparation of Ligands and Gold Nanoparticles
The gold nanoparticles (1)(2)(3)(4)(5) prepared in this work ( Figure 2) are constituted by an approximately 2 nm gold core coated with two components: simple monosaccharides (D-glucose or D-mannose) and DNJ derivatives. We played on the nature of the inner monosaccharide and varied the density of DNJ heads on the periphery of the nanoparticles. Nanoparticles 6 and 7, in Figure 2, bearing 100% monosaccharides with a short and linear five carbons (C 5 ) aliphatic linker, were prepared as control systems. For AuGNPs 1-5, the carbohydrate moieties guarantee their water dispersibility, while the water soluble and flexible amphiphilic linker was designed to protrude the DNJ heads from the glyco-coated nanoparticle core. In addition, the N-C 6 aliphatic linker of the active component DNJ is taking part in the interaction with the catalytic site entry [11,24]. This strategy of ligand presentation was based on our previous experience with similar constructs [21,22,25]. In particular, AuGNPs 6 and 7 were prepared employing, as exclusive components, 5-mercaptopent-1-yl β-D-glucopyranoside (βGlcC 5 SH, 13) or 5-mercaptopent-1-yl α-Dmannopyranoside (αManC 5 SH, 14), following well established protocols [25]. AuGNPs 1 and 2 were prepared by employing the monovalent DNJ derivative 12 as the active component and 13 or 14, respectively, as the inner component (Scheme 1). Scheme 1. Synthesis of the glyco-gold nanoparticles 1 and 2 decorated with the DNJ monovalent ligand 12.
The preparation of the monovalent DNJ derivative 12 is described in Scheme 1. The peracetylated azido-functionalized DNJ 8 [26] was reduced to the corresponding amino derivative 9 through a standard Staudinger reaction (PPh 3 , H 2 O, THF, reflux for 16 h) in 80% yield. The same reaction performed with polymer-bound PPh 3 [22,27], and refluxing in THF/H 2 O, required longer times and provided 9 in lower yields (48%).
The aminoalkyl DNJ 9 was then coupled to the thiol-protected carboxylic acid 10, prepared, as reported in the literature [21], using 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) and 1-hydroxybenzotriazole (HOBt) in dry DMF. The amide 11 was obtained in 67% yield after purification by flash column chromatography (FCC). Removal of the acetyl groups of 11 was performed under Zemplén conditions (MeONa) but using CD 3 OD, instead of MeOH, as solvent to follow the reaction via 1 H NMR. Once compound 12 was formed, it was directly used for the next step, avoiding its oxidation to disulfide. AuGNPs 1 and 2 were prepared by mixing 12 with βGlcC 5 SH (13) or αManC 5 SH (14), respectively, in the appropriate ratio (40/60) and then adding an Au(III) solution (the total amount of thiols being 3 equiv. with respect to gold) and the reducing agent NaBH 4 in excess (Scheme 1). The proportion of the ligands on the gold surface was evaluated by integrating diagnostic signals in the 1 H NMR spectrum of the initial mixture and in the 1 H NMR spectrum of the supernatant after AuGNPs formation (see Supplementary Materials for 1 H NMR spectra). The 40/60 active/inner component ratio allowed to preserve nanoparticle dispersibility in water [22]. After shaking for 2 h at room temperature, the supernatant was removed, and the nanoparticles were washed with methanol. The residue was dispersed in milliQ water, purified by dialysis, and freeze-dried. These nanoparticles were re-dispersed in water after the freeze-drying process, and no flocculation was observed even after several months. Full characterization with different techniques (see below) was carried out before enzymatic evaluation.
To further increase the DNJ loading onto AuGNPs, we combined this strategy with the dendron approach by preparing the trivalent DNJ derivative 18 (Scheme 2). The peracetylated azido-functionalized DNJ 8 was first reacted with the trialkyne 15 [28] to afford the dendron 16 in 85% yield through copper (I) catalyzed azide alkyne cycloaddition (CuAAC) reaction [29][30][31]. Subsequently, introduction of the linker was achieved by coupling the carboxylic acid 10 with 16 in the presence of 3-(diethoxyphosphoryloxy)-1,2,3-benzotriazin-4(3H)-one (DEPBT) as coupling agent affording the final ligand 17 in an excellent 94% yield. In analogy with the incorporation of the monovalent ligand 12 onto AuGNPs, trivalent thiol 18 was generated in situ, and its density on the nanoparticles was modulated with βGlcC5SH (13) (to yield AuGNPs 3) or αManC5SH (14) (to yield AuGNPs 4 and 5), as described above. The mannose-based AuGNPs 4 and 5 shared the same inner component but in different percentages (80% in 4, 60% in 5). AuGNPs 1-5 were subjected to quantitative 1 H NMR (qNMR) analysis in D 2 O with 3-(trimethylsilyl)propionic-2,2,3,3-d 4 acid (TSP-d 4 ) as an internal standard, following established protocols [21,32,33]. As an example, Figure 3a shows the qNMR for AuGNP 3. In this way, it was possible to quantify the DNJ amount on the nanoparticle by integration of a diagnostic signal (see Supplementary Materials and Table S1) with respect to the internal standard. An average gold core size in 1.6-2.1 nm range was determined by TEM imaging for the AuGNPs (Figure 3c,d) [25,34]. This size was also confirmed by UV-Vis spectra, which did not show an absorption maximum at around 520 nm, typical of gold nanoparticles with a bigger core diameter [35] (Figure 3b).

Biological Evaluation
The DNJ-coated AuGNPs 1-5 were evaluated against Jack bean α-mannosidase (JBαman), the glycosidase which showed the largest response to multivalent inhibitor presentation to date [3][4][5][6][7][8][9]. The inhibition mode and constants (K i ) were obtained from Dixon or Lineweaver Burk plots for a scale of inhibitor concentrations reflecting the DNJ concentration directly. The inhibition potency, relative to the corresponding monovalent reference 19 [36] (Figure 4), directly gives the relative potency per inhitope (rpn) ( Table 1).  The whole set of newly prepared heterovalent AuGNPs 1-5 showed good to excellent inhibitory activity towards JBα-man, with K i values in the low micromolar range to the nanomolar one. All behaved as competitive inhibitors (see Supplementary Materials). Conversely, β-D-glucoand α-D-mannoside decorated AuGNPs 6 and 7 did not inhibit the enzyme at 0.5 mg mL −1 (This concentration was a good compromise to allow spectrophotometric measurements due to the brown color of AuGNP solutions. The 12-valent mannosylated fullerene from Vincent and Nierengarten et al. (see ref [13]) was active at the hundreds of µM). However, trimerization of DNJ at the gold surface had a major impact: it strongly enhances the affinity towards JBα-man (3 vs. 1, 4 vs. 2, Table 1), with the affinity enhancement per inhitope increasing up to 1840 for AuGNP 4. A further enhancement of the inhibitory properties was reached upon increasing the loading of the trivalent DNJ derivative 18 onto the gold surface from 20% to 40%, with a K i at 84 nM and an rpn value of 3833 observed for AuGNP 5 (Table 1).
Noticeably, the dendron strategy is a double-edged sword. On the one hand, the intrinsic hindrance of the iminosugar tripod seems to prevent loading as high as linear arms (It is worth noting that for a same percentage loading of D-mannose of 60%, the DNJ concentration on AuGNP 5 is not 3 times higher than that of AuGNP 2, suggesting that the total loading on the particle is lower than for AuGNP 2. However, the ratio between iminosugar DNJ derivatives and monosaccharides still reflects the composition of the preparation mixture, as attested by 1 H-NMR spectra recorded before and after (supernatant) the formation of the AuGNPs (see Supplementary Materials)). On the second hand, although the DNJ concentration is almost identical (503 vs. 467 for AuGNP 2 and 5), the inhibition constant of AuGNP 5 is more than 100 times better (Table 1 and Table S1). Despite this lower coating, the multivalent effect, as judged by the rpn, is remarkably increased by two orders of magnitude. The spatial distribution of DNJ at the AuGNP 5, as obtained through the dendron strategy, is highly beneficial to the enzymatic affinity. AuGNPs thus represent an opportunity to compare multivalent systems, having similar concentrations in DNJ inhitopes, but with different local densities. In DNJ dendron-coated AuGNPs, the higher local concentration may favor the bind-and-recapture of the reversible inhibitor heads.

General Experimental Procedures for the Syntheses
Commercial reagents were used as received. All reactions were carried out under magnetic stirring and monitored by TLC on 0.25 mm silica gel plates (Merck F254). Column chromatographies were carried out on Silica Gel 60 (32-63 µm) or on silica gel (230-400 mesh, Merck, Darmstadt, Germany). Yields refer to spectroscopically and analytically pure compounds unless otherwise stated. 1 H NMR spectra were recorded on a Varian Gemini 200 MHz, a Varian Mercury 400 MHz, or on a Varian INOVA 400 MHz instrument at 25 • C. Additionally, 13 C NMR spectra were recorded on a Varian Gemini 200 MHz or on a Varian Mercury 400 MHz instrument. Chemical shifts are reported relative to CDCl 3 ( 13 C: δ = 77.0 ppm), or to CD 3 OD ( 13 C: δ = 49.0 ppm). The following abbreviations were used to designate multiplicities: s = singlet, d = doublet, t = triplet, m = multiplet, br = broad, and dd = double-doublet. Integrals are in accordance with assignments, coupling constants are given in Hz. For detailed peak assignments, 2D spectra were measured (COSY, HSQC). IR spectra were recorded with a IRAffinity-1S SHIMADZU system spectrophotometer. ESI-MS spectra were recorded with a Thermo Scientific™ LCQ fleet ion trap mass spectrometer. Elemental analyses were performed with a Thermo Finnigan FLASH EA 1112 CHN/S analyzer. Optical rotation measurements were performed on a JASCO DIP-370 polarimeter. TEM analysis was performed with a LaB6-TEM of type JEOL JEM-1400PLUS (40 kV-120 kV, HC pole piece) equipped with a GATAN US1000 CCD camera (2 k × 2 k).

Synthesis of Trivalent DNJ Derivative 16
To a solution of 8 (63 mg, 138 µmol) in 3 mL of THF/H 2 O = 2:1 CuSO 4 (2.1 mg, 13 µmol), sodium ascorbate (5 mg, 25 µmol) and 15 [28] (9.9 mg, 42 µmol) were added. The reaction mixture was stirred in microwave at 80 • C for 45 min, until a TLC analysis (CH 2 Cl 2 /MeOH 10: 1) showed the disappearance of the starting material (Rf = 0.46) and the formation of a new product (Rf = 0.00). The solvent was removed under reduced pressure and the crude was purified by FCC (CH 2 Cl 2 /MeOH/NH 3  To a solution of 11 or 17 in CD 3 OD (10 mg/mL), 30 equivalents of NaOMe were added, and the reaction mixture was left stirring for 2 h at 25 • C under nitrogen atmosphere. The complete disappearance of the starting material was attested via 1 H NMR, and the crude, containing compounds 12 or 18, respectively, was directly used for the preparation of AuGNPs.

General Procedure for the Preparation of AuGNPs 1-5
An aqueous solution of HAuCl 4 (25 mM, 1 equiv.) was added to a 12 mM methanolic solution of a suitable mixture of thiol-ending monosaccharide and DNJ ligands (3 equiv. overall). An aqueous solution of NaBH 4 (1 M, 27 equiv.) was then added in four portions, with vigorous shaking. The black suspension formed was shaken for 2 h at 25 • C. The residue was washed several times with MeOH. In order to effectively separate the nanoparticles from the supernatant, centrifugation (12,000 rpm, 2 min) was performed. The residue was dissolved in a minimal volume of HPLC gradient grade water and purified by dialysis (SnakeSkin ® Pleated Dialysis Tubing, 10,000 MWCO and Slide-A-Lyzer ® 10K Dialysis Cassettes, 10,000 MWCO). DNJ-coated AuGNPs were obtained as a dark-brown powder after freeze-drying and characterized via 1

Biological Evaluation
p-Nitrophenyl-α-D-mannopyranoside and α-mannosidase (EC 3.2.1.24, from Jack Bean, Km = 2.0 mM pH 5.5) were purchased from Merck (Darmstadt, Germany). Inhibition constants were determined by spectrophotometrically measuring the residual hydrolytic activities of the mannosidase against p-nitrophenyl-α-D-mannopyranoside in the presence and absence of the inhibitor with a VersaMax Microplate Reader. All kinetics were performed at 25 • C and started by substrate addition (20 µL) in a 100 µL assay medium (acetate buffer, 0.2 M, pH = 5) containing α-mannosidase (0.015 U/mL), in presence or absence of various concentrations of inhibitor and substrate (concentrations from km/8 to 2 km). After 15-30 min incubation, the reaction was quenched by the addition of 1M Na 2 CO 3 100 µL). The absorbance of the resulting solution was determined at 405 nm. Under these conditions, the p-nitrophenolate released led to optical densities linear with both reaction time and concentration of the enzyme. K i values were determined, in duplicate or triplicate, using the Dixon or Lineweaver-Burk graphical methods with Microsoft Excel [37].

Conclusions
In conclusion, we report, in this work, the first example of gold nanoparticles decorated with the iminosugar deoxynojirimycin (DNJ). Jack bean α-mannosidase (JBα-man) was chosen as the target enzyme for the biological evaluation due to its well-known responsiveness to the multivalent presentation of iminosugar inhitopes. Monovalent and trivalent DNJ-derivative functionalized with a thiol were synthesized and used in mixture with a glucose or mannose thiol-ending inner component for the in-situ preparation of heterovalent AuGNPs. AuGNPs, bearing only the sugar component, were also prepared as control systems. Biological assays towards JBα-man revealed a competitive inhibition for the whole set of AuGNPs 1-5, while glyco-coated AuGNPs 6-7 were inactive, thus demonstrating the fundamental role played by the iminosugar moiety for the inhibition. Comparison with the monovalent reference 19 highlighted affinity enhancements, per inhitope higher than 1600, for the AuGNPs 3-5 decorated with the trivalent DNJ-derivative, with the best result obtained for AuGNP 5 (K i = 84 nM, rpn = 3833). Interestingly, AuGNPs 2 and 5 share similar concentration in DNJ but vary by their local inhitope density, which makes them powerful tools to measure the impact of DNJ distribution onto the affinity enhancements. The significant gain of affinity, observed between the DNJ-coated and DNJ dendron-coated AuGNPs, highlights the importance of the bind-and-recapture effect in the complex, interconnected mechanisms underlying the inhibitory multivalent effects. Further investigation with these new nanosystems is currently ongoing in our laboratories.